Open source, MCP · 2025 to present

AI assistants had no reliable way to read food regulations across markets

Created and maintained by Ansuj Joshi, open source, 2025 to present

Food Regulatory MCP: ask a labelling rule in plain English

Situation

Regulatory data for food products is spread across Codex and EU Health Claims databases in formats an AI assistant cannot query reliably.

What we built

An open-source MCP server that gives AI agents structured, tool-based access to food regulatory data across 17 markets.

What changed

Works with Claude, ChatGPT, Cursor, Windsurf and GitHub Copilot, and is live at mcp-food-regulatory.vercel.app.

The problem

Asking a chat model about health-claim rules in a given market gets a confident answer and no source. Regulatory affairs teams need the actual record.

What we built

  • An MCP (Model Context Protocol) server exposing typed tools over structured regulatory databases.
  • Coverage of 17 markets, including Codex and EU Health Claims data.
  • Deployed on Vercel with FastMCP; installable in any MCP-capable client.

How it works

  • The assistant calls a tool with a market and a query; the server returns the structured record, so the answer cites data rather than memory.

What changed

17

markets covered

Source: Ansuj Joshi resume, 2026

  • Works with Claude, ChatGPT, Cursor, Windsurf and GitHub Copilot.

Stack

FastMCPPythonVercel

If this looks like your situation

Tell us where it hurts. We will say whether it is worth automating, and what we would build first.

Talk to usBook a call

Related work

FireOS compliance dashboard: statutory cycles tracked, a filing with an open defect blocked

FireOS, compliance operations

2026

Fire-safety compliance was running on spreadsheets, phone calls and memory

Statutory cycles are tracked as a system rather than remembered, and a filing that fails the rules cannot be submitted.

Built by Vyzma. Led by Vandan.

Agriva home page; the search box accepts plain-English queries

Agriva, real estate

2026

Buyers could not search farmland the way they talk about it

Chat is now the primary way buyers search on agriva.ai, and every answer is grounded in listing data rather than guessed.

Built by Vyzma. Led by Ansuj.

Agriva listing: seller text tidied, facts measured from map data

Agriva, real estate

2026

Brokers pasted WhatsApp text and phone photos, and that went live

Listings go live in English and local languages from one structured call, cheaper and more consistent than a call per language.

Built by Vyzma. Led by Ansuj.

← All work